Generate pages for nested files, not just top-level ones - #33
Merged
Conversation
Both "tree" and "tree_recursive" (at the top-level package scope, and the per-file scope on a directory entry) are misleadingly named - neither actually recurses. "tree" is filtered to top-level entries; "tree_recursive" is the exact same unfiltered top-level list (a no-op filter, since top-level entries never contain a path separator); and per-file `tree`/`tree_recursive` both just return one level of a directory's immediate children (treeFileGetTreeRecursive literally just returns `fs` unchanged). There was no existing way to get a real flat, full-depth listing of a repo's files. Add `flattenFiles` (Types.hs), which actually recurses into FolderContents and collects only file/blob leaves, and expose it as a new top-level `blobs` field in Repositories.hs's `package` scope. Left the existing `tree`/`tree_recursive` fields untouched, since other templates (and the readme/license lookup, which searches `tree`) may depend on their current, if misleadingly-named, behavior. Each entry's `path` in the flattened list is already the full path relative to the repo root - the tree walk in Repositories.hs accumulates this via prependParent as it descends, regardless of how deep a file is nested. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012RfZTyHvxLyVp4yaWmNvhN
processRepo' only ever called the file-page generator on the top-level tree (in both the force and non-force/getUpdatedFiles branches), never on anything reachable through a directory's FolderContents. So gitja has never actually written a file.html-equivalent page for anything nested in a subdirectory, independent of the "blobs" field or any template - the page itself was never generated, regardless of what a template linked to. Flatten tree with flattenFiles before both generation paths, so a file's depth no longer determines whether it gets a page. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012RfZTyHvxLyVp4yaWmNvhN
m-col
changed the title
m-col
pushed a commit
to m-col/mcol.xyz
that referenced
this pull request
Aug 30, 2026
gitja never generates pages for files nested in subdirectories, and top-level "tree" only ever contains top-level entries (see m-col/gitja#33, which also fixes this on gitja's side - needs a new gitja release to take effect here, since deploy.yaml downloads a release binary rather than building from source). Once that's out, this switches to gitja's new flattened "blobs" field, listing every file at its full path with no directory hierarchy - there's nothing for a directory entry to link to anyway, so this is simpler than the earlier dead-link workaround it replaces. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012RfZTyHvxLyVp4yaWmNvhN
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012RfZTyHvxLyVp4yaWmNvhN
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012RfZTyHvxLyVp4yaWmNvhN
Confirmed by reading the code: at both scopes (top-level package, and per-file on a directory entry), tree_recursive was identical to tree. Top-level tree's filter (notElem pathSeparator ...) was a no-op, since the list it filters only ever contains top-level entries in the first place - nested entries live inside each directory's own FolderContents, never flattened into it. Per-file, treeFileGetTreeRecursive just returned `fs` unchanged - no recursion despite the name. The docs described tree_recursive as "a list of *all* of the repository's/directory's contents", which was never actually true. "tree" is git's own name for a tree object's entries (a git tree is not itself recursive - that's exactly what it means for git to have separate tree objects per directory), so keeping that name for "this tree's entries" is accurate, not a simplification. Dropped the now-pointless filter/atTop logic along with tree_recursive, and updated DOCUMENTATION.md's Scopes/Attributes tables to match - "blobs" is documented there too, since it existed but wasn't previously listed. This doesn't add per-tree-entry pages for every object type a tree can contain (blob/tree/commit-as-submodule) - just removes a redundant, inaccurately-documented field. That's real future work, not this. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012RfZTyHvxLyVp4yaWmNvhN
blobs (leaves only) and tree_recursive (every entry, blobs and trees alike, at any depth) are different things - collapsing tree_recursive into blobs lost that distinction. Add flattenTree (Types.hs), the tree-and-blob equivalent of flattenFiles - keeps every entry including directories themselves as it descends, equivalent to `git ls-tree -r -t`. Used for both the top-level "tree_recursive" (package scope, Repositories.hs) and the per-file "tree_recursive" on a directory entry (treeAsLookup, Types.hs). flattenTree lives in Types.hs rather than alongside flattenFiles in Repositories.hs, since it's needed by treeAsLookup there too and Repositories.hs already depends on Types.hs (the reverse would be a cycle). Updated DOCUMENTATION.md's Scopes/Attributes tables to describe both tree_recursive and blobs accurately. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012RfZTyHvxLyVp4yaWmNvhN
Third special repo/ template, alongside commit.html and the newly-renamed blob.html: tree.html, generated once per tree (directory) object at any depth, mirroring how blob.html/commit.html already work per blob/commit. Uses flattenTrees (Repositories.hs) to walk every tree object including itself, and treeHref for its output path. Unlike blobs, a tree's own path never appears in a commit diff, so there's no cheap way to tell whether a tree needs regenerating from newCommits alone - tree pages are always regenerated on every run rather than incrementally, which is simpler and correct rather than risking the exact kind of silent staleness bug the blob-generation fix earlier addressed. "file" -> "blob" throughout (envFileTemplate/fileT/fileHref/fileDir -> envBlobTemplate/blobT/blobHref/blobDir, output moves from file/ to blob/), matching git's own object vocabulary and freeing up the name "file" from colliding with anything tree-related. Updated the three bundled templates (base/docs/stagit) so this doesn't silently break them: file.html -> blob.html (dropping their now-truly- unreachable is_directory branches, since blobs are never directories), the existing once-per-repo file-listing tree.html -> files.html (freeing the name for the new per-object template), and a real tree.html per-object page for each, built from what those old is_directory branches were already trying to do. Every href referencing the old file/ output directory is updated to blob/, and files.html/tree.html's child links now route to tree/ or blob/ depending on the entry's own type. Updated DOCUMENTATION.md's folder structure, scopes and attributes tables to describe blob/tree/commit as the three special per-object templates, and the blob/tree attribute table entries as shared (most attributes apply to both, a few are n/a on one side). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012RfZTyHvxLyVp4yaWmNvhN
m-col
pushed a commit
to m-col/mcol.xyz
that referenced
this pull request
Aug 30, 2026
Matches m-col/gitja#33 (not yet released): file.html -> blob.html, the file-listing tree.html -> files.html, and a real per-tree tree.html so subdirectories actually get their own page once that release lands. Every /code/<repo>/file/... link updates to /code/<repo>/blob/... - including the qtools blog post link fixed earlier this session for the same reason. This won't do anything until a new gitja release exists (deploy.yaml downloads gitja's latest release binary rather than building from source) - until then /code will keep working exactly as it does now, since the currently-released gitja binary doesn't know about blob.html or the new tree.html and just ignores them. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012RfZTyHvxLyVp4yaWmNvhN
m-col
pushed a commit
to m-col/mcol.xyz
that referenced
this pull request
Aug 30, 2026
Matches m-col/gitja#33 (not yet released): file.html -> blob.html, the file-listing tree.html -> files.html, and a real per-tree tree.html so subdirectories actually get their own page once that release lands. Every /code/<repo>/file/... link updates to /code/<repo>/blob/... - including the qtools blog post link fixed earlier this session for the same reason. Sequencing matters here: the currently-released gitja binary looks for "file.html" by exact name. Once this merges and deploy.yaml runs against that old binary, it won't find file.html (renamed to blob.html) and will silently stop generating individual blob/file content pages entirely - the files.html listing and commit pages keep working fine (they're unaffected), but every "view this file's contents" link will 404 until gitja#33 is merged, released, and picked up by the next deploy. Consider holding off merging this until that release exists, to avoid the gap. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012RfZTyHvxLyVp4yaWmNvhN
Fixes the CI failure on the previous commit: test.sh's golden-file test still expected output under gitja/file/... (the pre-rename output directory), and test/templates/repo/file.html was never updated to match the blob rename. Renamed that test template to foreach.blob.html (see below), updated its dumped "file"/"scope: file" labels to "blob"/"scope: blob" (the values themselves - paths, hrefs, contents - are unchanged, since it's the same underlying object under a renamed scope), moved test/expected/gitja/file/ to test/expected/gitja/blob/ with matching label updates, and fixed test.sh's TESTS array paths. Also rename the three special per-object templates to foreach.blob.html/foreach.commit.html/foreach.tree.html (from blob.html/commit.html/tree.html), across the bundled templates, the test template, and Env.hs's lookups. Makes it visually obvious in a directory listing which templates in repo/ run once per repo vs. once per object - previously "blob.html"/"commit.html"/"tree.html" looked like just three more arbitrarily-named repo templates alongside index.html/refs.html/etc, despite being the special-cased ones. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012RfZTyHvxLyVp4yaWmNvhN
m-col
pushed a commit
to m-col/mcol.xyz
that referenced
this pull request
Aug 30, 2026
Matches m-col/gitja#33: blob.html/commit.html/tree.html -> foreach.blob.html/foreach.commit.html/foreach.tree.html, so it's visually obvious which templates in repo/ run once per object vs. once per repo. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012RfZTyHvxLyVp4yaWmNvhN
… field TreeEntry better reflects that these represent any tree object (blob, tree, or submodule), and entries/trees give the flattened lists clearer, distinct names.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes gitja never generating pages for nested files, adds a
foreach.tree.htmlper-object template (alongsideforeach.blob.htmlandforeach.commit.html) so subtrees get their own generated page, and renamesTreeFile→TreeEntry/tree_recursive→entriesfor clarity, plus a newtreesfield listing every directory in the repo.